kubectl top pod — CPU/Memory Usage & One-Liners¶

Reference: https://kubernetes.io/docs/reference/kubectl/cheatsheet/

See also: node_resource.md.md for kubectl top node, millicores, --sort-by, --no-headers, and the full shell pipeline reference.

The Exercise¶

Find the pod that consumes the most CPU across all namespaces. Store the result in high_cpu_pod.txt in the format: pod_name,namespace

Solution:

kubectl top pod --sort-by=cpu -A --no-headers | head -1 | awk '{print $2 "," $1}' > high_cpu_pod.txt

Full Breakdown — Every Part¶

Stage 1: `kubectl top pod --sort-by=cpu -A --no-headers`¶

Flag	What it does
`--sort-by=cpu`	Sort output by CPU usage, highest first. Also accepts `memory`
`-A`	Short for `--all-namespaces` — show pods from every namespace including `kube-system`
`--no-headers`	Suppress the column header row so `head -1` gets the first data row, not the header

Output:

kube-system   kube-apiserver-controlplane   98m    282Mi
kube-system   etcd-controlplane             22m    58Mi
default       my-app-pod                    15m    120Mi
kube-system   coredns-abc123                5m     18Mi

Stage 2: `| head -1`¶

Takes only the first line. Since output is sorted by CPU descending, first line = highest CPU pod.

kube-system   kube-apiserver-controlplane   98m    282Mi

Stage 3: `| awk '{print $2 "," $1}'`¶

awk extracts fields by whitespace position.

Expression	Value
`$1`	`kube-system` (namespace)
`$2`	`kube-apiserver-controlplane` (pod name)
`$2 "," $1`	`kube-apiserver-controlplane,kube-system`

The required format is pod_name,namespace — so $2 comes first, then a literal comma ",", then $1. Reversed from the column order.

Stage 4: `> high_cpu_pod.txt`¶

Writes the output to the file. Creates if it doesn't exist, overwrites if it does.

Final result in the file:

kube-apiserver-controlplane,kube-system

The #1 Mistake — Column Positions Change With -A¶

This is the most common error on this type of question. The namespace column gets prepended when you use -A.

kubectl top pod --no-headers (single namespace):

my-pod   15m   120Mi
$1       $2    $3

$1 = pod name

kubectl top pod -A --no-headers (all namespaces):

kube-system   kube-apiserver-controlplane   98m   282Mi
$1            $2                            $3    $4

$1 = namespace, $2 = pod name

If you use awk '{print $1}' on -A output, you get the namespace, not the pod name. The answer is wrong and you won't know why unless you know this.

Compare with top node which never has a namespace column:

controlplane   105m   10%   1292Mi   60%
$1             $2     $3    $4       $5

$1 = node name always.

Common Exam Variants¶

Highest CPU pod, all namespaces → pod_name,namespace:

kubectl top pod -A --sort-by=cpu --no-headers | head -1 | awk '{print $2 "," $1}' > result.txt

Highest memory pod, all namespaces → pod_name,namespace:

kubectl top pod -A --sort-by=memory --no-headers | head -1 | awk '{print $2 "," $1}' > result.txt

Highest CPU pod in a specific namespace → just pod name:

kubectl top pod -n kube-system --sort-by=cpu --no-headers | head -1 | awk '{print $1}' > result.txt

No -A here, so $1 is the pod name again.

Highest CPU node → node name:

kubectl top node --sort-by=cpu --no-headers | head -1 | awk '{print $1}' > result.txt

Current context + highest memory node → context,node:

echo "$(kubectl config current-context),$(kubectl top node --sort-by=memory --no-headers | head -1 | awk '{print $1}')" > result.txt

Verify¶

Always cat the file after writing:

cat high_cpu_pod.txt

Check: one line, comma-separated with no spaces, correct order (pod_name,namespace not namespace,pod_name).

Quick Reference¶

# Pods — current namespace ($1=pod name)
kubectl top pod
kubectl top pod --sort-by=cpu --no-headers

# Pods — all namespaces ($1=namespace, $2=pod name)
kubectl top pod -A
kubectl top pod -A --sort-by=cpu --no-headers

# Nodes ($1=node name)
kubectl top node --sort-by=memory --no-headers

# One-liners
kubectl top pod -A --sort-by=cpu --no-headers | head -1 | awk '{print $2 "," $1}'    # pod_name,namespace
kubectl top pod -A --sort-by=memory --no-headers | head -1 | awk '{print $2 "," $1}' # pod_name,namespace
kubectl top node --sort-by=cpu --no-headers | head -1 | awk '{print $1}'              # node name only